The Presolution Period and the Detection of Statistical Associations
The Presolution Period and the Detection of Statistical Associations
复制标题
预测周期和统计关联的检测
DOI：
--
复制
发表时间：
1983
期刊：
影响因子：
--
通讯作者：
E. G. Heinemann
中科院分区：
文献类型：
--
作者：
E. G. Heinemann
当研究辨别力的获得时，人们经常会发现一个初始阶段，在此期间，所研究的因变量没有可测量的变化（例如，正确的选择）。这一时期通常被称为“前溶解期”（PSP）（Krechevsky，1932）。关于前溶解期，人们做了许多实验，以决定是采用不连续性理论（拉什利，1929年;克列切夫斯基，1932年，1938年）还是采用连续性理论（斯宾塞，1936年，1945年），大多数实验都是通过研究前溶解期歧视的逆转对习得过程的影响来进行的。这项研究的评论者（古德里奇、罗斯和瓦格纳，1961年;萨瑟兰和麦克米伦，1963年）都认为，如果辨别刺激刺激了适当的受体器官，那么在前溶解期的辨别逆转会延迟反向辨别的获得。尽管围绕间断-连续之争进行了大量的研究工作，但对决定解前期长度的变量却知之甚少。可以定量预测这种效应的理论还没有发展出来。这一章的目的是提出一个定量模型的过程中发生的presolution期间，连同一些相关的数据。该模型的目的是适用于一种特定类型的实验情况下，即，一个离散的试验实验中，两个刺激值之一，从一个连续的，如声音强度，是在每次审判。受试者在刺激之一S1的存在下做出一个反应R1，在另一个刺激S2的存在下做出另一个反应R2，从而得到奖励。1.在本章所述的实验中，被试为白色白鸽，刺激S1和S2为白色噪声强度，R1和R2为在标准鸽子训练室中啄两个反应键。理论治疗将动物视为统计关联的检测器，该统计关联已在刺激强度水平、与所做出的响应相关联的刺激以及响应的后果之间进行了安排。正如信号识别理论中通常所做的那样，假设在无限重复的呈现中，两个刺激强度S1和S2产生的感觉效应是正态分布的，具有不同的均值μl和μ2，以及共同的标准差σ。进一步假设，在一系列的试验中，受试者记住了R1之后是奖励时存在的感觉效果，以及R2之后是奖励时存在的感觉效果。受试者的任务是决定是否平均感觉效应，M1和M2，这之前奖励R1和R2的反应不同可靠。如果它们确实存在可靠的差异，那么所考虑的刺激可以预测反应的结果，并且将它们用作反应选择的基础是有益的。本理论的核心是一个程序的正式模型，受试者使用该模型来决定是否认为M1和M2之间的差异是显著的。更具体地说，受试者检验假设H 0，
When studying the acquisition of discriminations, one often finds an initial period during which there is no measurable change in the dependent variable under study (e.g., percent correct choices). This period is usually called the “presolution period” (PSP) (Krechevsky, 1932). Numerous experiments on the presolution period have been done to decide between discontinuity theory (Lashley, 1929; Krechevsky, 1932, 1938) and continuity theory (Spence, 1936, 1945), most often by studying the effect that the reversal of the discrimination during the presolution period has on the course of acquisition. Reviewers of this work (Goodrich, Ross, and Wagner, 1961; Sutherland and Mackintosh, 1963) agree that discrimination reversal during the presolution period retards acquisition of the reversed discrimination, provided that the discriminative stimuli stimulate the appropriate receptor organs. In spite of the large research effort organized around the discontinuity-continuity controversy, relatively little is known of the variables that govern the length of the presolution period. Theories from which such effects may be predicted in a quantitative way have not been developed. It is the purpose of this chapter to present a quantitative model for the processes occurring during the presolution period, together with some relevant data. The model is intended to apply to a specific type of experimental situation, namely, a discrete-trial experiment in which one of two stimulus values from a continuum, such as sound intensity, is presented on each trial. The subject is rewarded for making one response, R1, in the presence of one of the stimuli, S1, and an alternative response, R2, in the presence of the other stimulus, S2. 1 In the experiments to be described in this chapter, the subjects were White Carneaux pigeons, the stimuli, S1 and S2, were levels of intensity of white noise, and R1 and R2 were pecks on two response keys in a standard pigeon training chamber. The theoretical treatment regards the animal as a detector of the statistical association that has been arranged among the stimulus intensity levels, the stimulation associated with the responses that are made, and the consequences of the responses. As is commonly done in signalrecognition theory, it is assumed that over indefinitely repeated presentations, the two stimulus intensities, S1 and S2, produce sensory effects that are normally distributed with different means, µl and µ2, and a common standard deviation σ. It is further assumed that over a series of trials, the subject remembers the sensory effects that were present when R1 was followed by reward and those present when R2 was followed by reward. The subject’s task is to decide whether the mean sensory effects, M1 and M2, which preceded rewarded R1 and R2 responses differ reliably. If they do differ reliably, then the stimuli under consideration predict the outcome of the responses, and it would be profitable to use them as a basis for response choice. The heart of the present theory is a formal model of the procedure that the subject uses to decide whether or not to consider the difference between M1 and M2 significant. More specifically, the subject tests the hypothesis, H0, that